In order to provide the framework with a quantitative basis, simulated values of the used indicators are compared with threshold values. To obtain these simulations, several (computer) models were used throughout this work. The choice for models is first made and motivated in Section 2.4.1. A brief outline of the working of these models is given in Section 2.4.2.
2.4.1 Model choice
As mentioned in Section 2.2, the framework does not prescribe the use of specific models, since the model choice strongly depends on the case’s objectives, the avail- ability of data and the availability of models. However, because of time constraints and limited model availability, it was decided to fix the choice of models across the cases treated in this work. It would be too time-consuming to use multiple models for each climate hazard. Moreover, most software packages can only be used with licenses, which were not always available.
Although the model choice has been fixed in this research, the framework itself is sufficiently generic to allow the use of other models which may fit the research ob- jectives better. The framework does not prescribe this but rather leaves this freedom to the user.
Pluvial flooding
To simulate the effects of pluvial flooding, several software packages were consid- ered. A wide variety of flood models is available, but the model here should at least be able to simulate sewage flows (1D) and combine this with overland flow and runoff (2D) to arrive at the desired generic indicator inundation depth. This limited the list of possible choices to only a few models, being 3Di, SOBEK2, D-HYDRO and InfoWorks ICM. Both SOBEK2 and InfoWorks ICM are used within Witteveen+Bos, but InfoWorks is used most often for modeling pluvial flooding and water hindrance. It was therefore decided to use InfoWorks ICM as a software package to simulate the effects of pluvial flooding, partly based on a practical motivation.
2.4. CLIMATE HAZARD SIMULATION TOOLS 25
Heat stress
As mentioned previously, heat stress modeling still is in its infancy. There are no dedicated software packages readily available and the quantitative basis for norms and model output is narrow. The discussion regarding standardization of heat stress modeling approaches has gained momentum in the Netherlands and is currently (2019) still ongoing.
Mirzaei (2015) distinguishes heat stress models according to their scale of appli- cation: building-scale, micro-scale and city-scale models. The former are limited to the heat stress in and around individual structures; the latter comprise entire cities. It here makes sense to choose a heat stress model at the micro-scale, suitable for a neighborhood. The scale of such micro-scale models is still too large to take into account air flows and turbulence, which leaves the option to make use of a simplified energy balance model.
Witteveen+Bos has co-developed their own heat stress model in collaboration with several institutes, called Urban Climate Assessment and Management (UCAM). The model is based on an energy balance and mainly takes into account land and material use and does not simulate fluid dynamics. It was specifically designed to assess heat stress at the neighborhood scale and its primary outcome is the quantification of the (daily-averaged) UHI effect. UCAM was the only available heat stress model fitting the neighborhood scale. Despite its limitations, it was decided to use UCAM here since it still gives a good impression of the heat stress experienced and shows potential vulnerable locations.
Groundwater
No groundwater computer models were used in this study due to time and practi- cal constraints. Instead it was decided to use the analytical methods described by Hooghoudt (1940) and Bear (1979), and Koppejan (1942) to assess high and low groundwater levels respectively. These are classical groundwater equations which form the basis of various groundwater models. For the scale of application (neigh- borhood), it was considered justifiable to use these equations here.
2.4.2 Model descriptions
InfoWorks ICM
InfoWorks is a platform of water-related simulation models consisting of InfoWorks ICM (Integrated Catchment Model), SD (Stormwater Drainage), RS (River Systems) and CS (Collection System) among others. These models can be used separately
but can also be easily integrated. The description here focuses on InfoWorks ICM as this model was used to obtain simulations of the sewage system’s behavior during and after precipitation events.
ICM has the possibility to integrate urban and river catchments in one model and is both a hydraulic and hydrologic model. It simulates the rainfall-runoff process in an urban environment through the allocation of subcatchment areas to drainage points (e.g. manholes). Within these subcatchments, a distinction is made between pervious and impervious areas (and their type, such as sloped or open or closed paved) in order to accurately model infiltration and runoff. For 1D infiltration, fixed characteristics for each subcatchment need to be specified. For 2D infiltration, ICM uses either (1) a fixed percentage of rainfall, (2) a constant infiltration rate based on saturated soil conditions or (3) a variant of the Horton Equation (Horton, 1933) over a 2D area.
Water which reaches the stormwater sewage is modeled using the 1D module of ICM; surface water flow is simulated using the 2D module. A more elaborate description of the modeling approach in InfoWorks ICM can be found in Appendix B.
UCAM
To simulate and assess local heat stress and air quality, the Urban Climate As- sessment and Management (UCAM) model was developed (Witteveen+Bos, 2014; Tijdschrift Milieu, 2015). It was co-developed by Witteveen+Bos, Wageningen UR and the KNMI. The discussion about further standardization of heat stress modeling approaches and output to ensure better comparability among assessed areas is still ongoing.
UCAM’s working is based on the effects of the heatwave that struck the Nether- lands between July 14 and July 19, 2006 on a number of standard Dutch neighbor- hoods. To this end, the weather characteristics during this heat wave were used as input in the Weather Research and Forecasting (WRF) model (for a comprehensive overview, see Powers et al. (2017)) to simulate the energy balance discussed below. It was configured to account for small urban areas typical for the Netherlands and was validated using observations in the rural area Cabauw. The effects on the Ur- ban Heat Island (UHI) effect (the air temperature difference between urban and rural areas) of several physical neighborhood parameters (to be discussed subsequently) were monitored and the sensitivity of the UHI effect to these parameters was deter- mined. As a result, the model is able to approximate the extent of the UHI effect given that information about the relevant neighborhood parameters is available.
The model is governed by the following energy balance:
2.4. CLIMATE HAZARD SIMULATION TOOLS 27
Figure 2.2: Illustration of the energy budget (Eq. 2.1) as used in UCAM. Adapted from Akbari et al. (n.d.)
where Q is (solar) radiation [M T-3; usually W m-2], AH anthropogenic heat, H the
sensible heat in the area (UCAM’s UHI output),LE latent heat, andGthe storage of
heat in the area. These parameters are illustrated in Figure 2.2.
During a heat wave, the solar radiation Q is at its highest. The energy balance
in a city described by Eq. 2.1 differs from rural areas in the sense that (1) AH is
larger because of more human activities, (2) urban areas are developed and feature low albedo values and ample opportunities for heat to be stored in buildings and (3) the amount of greenery is generally limited in cities, restricting the latent heat. As a result, much of the excess heat will be stored in buildings and roads (G). This
storage term is eventually converted to sensible heat, which explains the strong UHI effect at night.
The model works with the Local Climate Zone classification system as described by Stewart and Oke (2012) for neighborhood types. The model differentiates be- tween LCZ2 (compact midrise buildings), LCZ3 (compact lowrise), LCZ5 (open midrise), LCZ6 (open lowrise) and LCZ23 (a combination of LCZ2 and LCZ3; similar to typical Dutch row houses). Compact here refers to a low number of trees, largely paved land cover and the use of concrete construction materials; open refers to an abundance of vegetation and pervious land cover.
Moreover, the model accounts for the albedo-effect of the built-up area. Albedo can be considered the extent of the reflection of radiation and the albedo number varies between 0 (no reflection, maximum absorption) to 1 (maximum reflection, no
absorption) and as such is often expressed as a percentage. The albedo numbers of roofs, facades and roads are considered separately. As shown in Equation 2.1, an- thropogenically generated heat contributes to the UHI, especially since urban areas often have many inhabitants. However, its effects are thought to be minor and it is difficult to approximate the degree of anthropogenic heat beforehand. It is therefore usually set to a default value. The amount of green surface area is another param- eter that is taken into account. Greenery does not store heat, but rather converts it to latent heat (LE in Eq. 2.1), thereby loweringH and the UHI effect.
UCAM’s primary input is the land use classified according to five categories, some with their own characteristic values. These categories are (parameter values bracketed): paved surface (road albedo), structures (roof and facade albedo and structure height), gardens (% unpaved/green), water and greenery. The study area is discretized into a grid and for each cell the model determines a LCZ based on the presence and relative position of these categories and structure height. The size of the entire area is also considered. The determination of LCZ requires semi-manual work of classifying the projected land use in the study area.
UCAM’s primary output is the daily-averaged UHI effect (in ◦C); the sensible
heat in the area (parameter H from Eq. 2.1). The daily average figure gives a
good impression of both the amplified heat during the day in urban areas and the (lack of) cooling during the night. Also, the duration for which people are exposed to amplified heat is usually longer than just a few hours. To arrive at this value, UCAM first calculated the maximum UHI in each grid cell, based on its LCZ and the characteristics of the 2006 heat wave according to the WRF model. It then compensates the maximum possible UHI effect with reduction factors. Reductions in UHI effect due to albedo effects and greenery are subtracted from the maximum UHI effect. For example, a grid cell that contains only paved surface and no structures or greenery, will show the maximum UHI effect possible as there are no reduction factors at play. The role of water in heat stress is somewhat ambiguous: water may absorb heat and cause cooling through evaporation. However, surface water becomes warmer as the summer progresses, causing its mitigating effect on heat stress to dissipate or even become adverse. While it is very difficult to predict or simulate the temperature at a certain location with good accuracy, UCAM seems to give reasonable values for the UHI effect. Successful modeling studies with UCAM have already been undertaken in Ghent, Almere and Gouda.
Because the modeling of heat stress is still underdeveloped and no national stan- dards exist, it is difficult to validate the model and assess a neighborhood with clear criteria. In an attempt to provide some assessment scale and allow for comparabil- ity across neighborhoods, UCAM works with the so-called heat-index. This index, which varies from 0 (no UHI) to 1 (large UHI effect; severe heat stress), is based on
2.4. CLIMATE HAZARD SIMULATION TOOLS 29
the RIVM norms for ozone concentration in the air. The heat-index is linearly related to the UHI and a heat index of 1.0 corresponds roughly with a UHI effect of 3.3 ◦C
(daily-averaged). This is the maximum value for the UHI effect that can be obtained in the simulations in a Dutch setting. This approach is based on the European WHO guidelines for ozone concentrations, because exposure to ozone peak values yields similar health risks as heat stress: both typically occur during heat waves and cause risks for the short term. The heat-index is used to formulate the final assessment in four categories: 0 through III. This classification is depicted in Table 2.1.
Table 2.1: UCAM heat stress categories according to the heat-index value
Category Heat-index Assessment Interpretation
0. 0 - 0.29 Comfortable No effects, no additional heat risks for the urban environment.
I. 0.3 - 0.57 Acceptable
Limited heat risks for the urban environment. Acceptable for a maximum of 25 days per year
II. 0.58 - 0.86 Risky
Additional heat risks for the urban environment. Comparable with the information threshold for ozone concentration.
III. 0.87 - 1.0 Unacceptable
Large additional heat risks for the urban environment. Comparable with
the alarm threshold for ozone concentration.
It is important to remark that the UHI effect values calculated by UCAM are daily- averaged. In reality, the local deviation in temperature from the rural surroundings may reach much higher values. This is especially the case toward the end of the day as the sun has set and rural areas cool down faster than urban areas.
The UCAM GIS tool is supplemented by a spreadsheet in MS Excel. This can be used to compose a graph of the experienced temperature at a specific location (based on LCZ and vegetation percentage, as determined in the GIS tool) over the course of a hot day. The graph can provide more insight in the temperature difference between developed and rural areas at specific times, rather than just a daily-averaged figure.
Groundwater hindrance
The analytical formula of Hooghoudt (1940) was used to estimate the differential head between drains in the neighborhood. The formula can be written as follows:
m0= −8·K2·d
p
(8·K2·d)−4·4·K1· −s·L2
2·4·K1 (2.2)
wherem0denotes the differential head between two drains [L],K the hydraulic con-
ductivity of soil layers 1 and 2 [L T-1],d the equivalent depth [L; see Eq. 2.3],s the
specific discharge to be drained [L T-1] andLthe distance between drains [L].
The equivalent depth d is a function of several other parameters, according to
the following equation:
d= D
1 + 8DπL ·ln(πrD) (2.3)
in whichDis the vertical distance between the drain and the impermeable layer (soil
layer 1) [L] andr is the wetted perimeter of the drain [L].
In cases where the drainage levels on both sides are not identical, the following equation can be used to approximate the groundwater level (Bear, 1979):
h2(x) =h20−h 2 0−h2L L x+ s K(L−x)x (2.4)
wherehis the groundwater level [L] as a function of the horizontal distance x [L] on
the domain between 0 (left drain) and L (right drain). The maximum groundwater level can then be found graphically.
Drought-induced soil subsidence
In addition to water hindrance resulting from high groundwater levels, dry spells may cause the groundwater table to become much lower and induce soil subsidence and subsequent damage to property (Corti et al., 2009; Swiss Reinsurance Company, 2011). To quantify the extent of this phenomenon in the study area, we used the formula of Koppejan (1942). It combines the previously reported expressions for soil subsidence by Terzaghi et al. (1967) and Keverlingh Buisman (1940) and account for the duration of the soil loading or lowering of the water table. It can be written as follows: S(t) = T 1 Cp + 1 Cs ·log(t) ·ln pi+ ∆pi pi (2.5) whereS is the subsidence [L],t is time [T],T is the thickness of the layer subject to
subsidence [L],Cpthe consolidation constant for direct effect [-],Csthe consolidation
constant for secondary effects [-] andpithe intergranular pressure in the considered
2.5. GENERIC FORCING 31
water pressure from the soil pressure at the bottom of the soil layer susceptible to subsidence.
Equation 2.5 can be used to estimate the subsidence of soils that are loaded, but also for soils where the water table is reduced as the effect on the intergranualar pressure is similar.